Machine learning for frequency estimation of power systems

In this paper the application of machine learning techniques for on-line dynamic security assessment of power systems is presented. Decision trees (DT), artificial neural networks (ANN) and entropy networks (EN) are developed and applied on the power system of Crete, the largest Greek island. Comparison of these methods reveals their relative advantages and disadvantages. These methods have been integrated in the dynamic security assessment module of the advanced control system of Crete island, helping to identify the operating conditions and parameters that lead to a less robust operation of the system. The results are considered very satisfactory, both in accuracy that increases the reliability of the method and in computational time, which is a necessity for real time applications.